Experience

  1. Research Assistant · Human-in-the-Loop VLA Learning

    Georgia Institute of Technology

    Supervised by Dr. Ye Zhao. Develop learning and control pipelines for dexterous clinical robotic manipulation, including high-precision pipetting.

    • Fine-tune GR00T, QwenOFT, and π0.5 policies with supervised fine-tuning and online actor–critic reinforcement learning with human intervention.
    • Improved real-robot millimeter-level task success from 10% to 85% through iterative hardware reinforcement learning.

  2. Research Assistant · E/I-Separated Recurrent Neural Networks

    Shanghai Qizhi Institute

    Supervised by Dr. Lu Mi. Investigated Dale’s principle as an inductive bias for recurrent networks, with experiments in multi-task learning, continual learning, and reinforcement learning.

    First-author paper accepted to NeurIPS 2026.

  3. Research Assistant · Inverse Reinforcement Learning

    BRAINML Lab, Georgia Institute of Technology

    Supervised by Dr. Anqi Wu. Developed a framework to discover reusable motor motifs and recover reward structure from behavioral trajectories. Applied expectation–maximization to multi-reward MDPs to infer latent behavioral intentions.

    First-author work on reusable motor motifs accepted to NeurIPS 2026; co-authored ICML 2025 work on switching, history-dependent rewards.

  4. Research Assistant · Reinforcement Learning Models of Replay

    Tsinghua University & New York University
    Supervised by Dr. Sen Song and Dr. Marcelo Mattar. Built a modular reinforcement learning model in JAX to study replay, planning, and cognitive-map learning. Analyzed neural manifolds and reproduced biological replay experiments.

Education

  1. PhD in Computer Science and Engineering

    Georgia Institute of Technology
    GPA: 4.0/4.0

  2. BS in Chemistry

    Peking University
    GPA: 3.70/4.00
Skills
Learning & Control
VLA fine-tuning, imitation learning & behavior cloning
Actor–critic RL, human-in-the-loop RL & inverse RL
Dexterous manipulation, humanoid control & world models
Engineering
Python, PyTorch, JAX, C & C++
NVIDIA GR00T, QwenOFT, π0.5 & StarVLA
Isaac Sim / PhysX, MuJoCo & Unitree G1
Codex & Claude CLI for research development
Selected Awards
National Scholarship (2/165) · Samsung Scholarship (20/160) · National Chemistry Olympiad Silver Medal